Understanding Clique Formation in Social Networks - An Agent-Based Model of Social Preferences in Fixed and Dynamic Networks

Pratyush Arya, Nisheeth Srivastava

Abstract: This paper presents results from in silico experiments trying to uncover the mechanisms by which people both succeed and fail to reach consensus in networked games. We find that the primary cause for failure in such games is preferential selection of information sources. Agents forced to sample information from randomly selected fixed neighborhoods eventually converge to a consensus, while agents free to form their own neighborhoods and forming them on the basis of homophily frequently end up creating balkanized cliques. We also find that small-world structure mitigates the drive towards consensus in fixed networks, but not for self-selecting networks. Preferentially attached networks appear to show the highest convergence to one color, thereby showing resilience to balkanization of opinion in self-selecting networks. We conclude with a brief discussion of the implications of our findings for the representation of behavior in socio-cultural modeling.